fix: sanitize hyphens in Cypher relationship names to prevent syntax errors

[ajatprabha]

Description

Adds hyphen (-) sanitization to sanitize_relationship_for_cypher() function to prevent Neo4j Cypher syntax errors when relationship names contain hyphens.

Problem: The sanitize_relationship_for_cypher() function in mem0/memory/utils.py handles 40+ special characters but was missing hyphen sanitization. This caused Cypher syntax errors when mem0's LLM extracted relationship names with hyphens from natural language (e.g., "non-compliant", "sub-category", "co-worker").

Error Example:
neo4j.exceptions.CypherSyntaxError: Invalid input '-': expected a parameter...
"MERGE (source)-[r:non-compliant]->(destination)"
^

Solution: Added "-": "_dash_" to the char_map following the existing naming convention (similar to "/": "_slash_", "+": "_plus_").

Impact: Common use cases with hyphenated terms now work correctly with graph memory enabled.

Fixes #3838

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Refactor (does not change functionality, e.g. code style improvements, linting)
• Documentation update

How Has This Been Tested?

Manual Testing:

1. Created a mem0 Memory instance with Neo4j graph store enabled
2. Added memory with natural language containing hyphens: "Mark all Y things as non-compliant"
3. Before fix: Cypher syntax error on MERGE ()-[r:non-compliant]->()
4. After fix: Relationship created successfully as non_dash_compliant

Test Configuration:

• Python 3.13
• Neo4j 5.x (via Docker)
• Qdrant vector store
• Azure OpenAI (gpt-5-mini)

Verified the sanitization follows existing patterns and doesn't break other character replacements.

• Test Script (manually reproduced the issue with real Neo4j instance)
• Unit Test (should be added in future PR)

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas (not needed - follows existing pattern)
• I have made corresponding changes to the documentation (not applicable)
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works (should be added separately)
• New and existing unit tests pass locally with my changes (no tests exist for this function yet)
• Any dependent changes have been merged and published in downstream modules (not applicable)
• I have checked my code and corrected any misspellings

Maintainer Checklist

• closes Bug: sanitize_relationship_for_cypher() doesn't handle hyphens causing Neo4j Cypher syntax errors #3838
• Made sure Checks passed

[CLAassistant]

All committers have signed the CLA.

[ajatprabha]

Testing Summary

Test Environment

• Setup: Local mem0 installation via uv pip install -e /path/to/mem0
• Stack: Qdrant (vector store) + Neo4j (graph store) + Azure OpenAI (gpt-5-mini)
• Test script: Custom memory spike with 8 demos covering various relationship extraction scenarios

Test Cases

1. Explicit Hyphenated Relationship Types

Forced the LLM to extract relationship types containing hyphens by using compound terms in natural language:

Test Case: "Acme Corp is a sub-contractor of Mega Industries for the Q4 project"

Before fix:
neo4j.exceptions.CypherSyntaxError: Invalid input '-': expected a parameter...
"MERGE (source)-[r:sub-contractor]->(destination)"
^

After fix:

# Successfully created relationship with sanitized type
MATCH ()-[r]->() WHERE type(r) = 'sub_dash_contractor'
# Returns: 1 relationship

2. Additional Test Scenarios

- "follow-up": "Invoice INV-002 is a follow-up to Invoice INV-001"
- LLM rephrased naturally to follows_up_on (no hyphen in output)
- "by-product": "Chemical XYZ is a by-product of manufacturing process ABC"
- LLM rephrased to is_by_product_of (no hyphen in output)

3. Entity Name Verification

Confirmed that entity names (node properties) with hyphens work correctly:
MATCH (n)-[r]->(m) RETURN n.name, type(r), m.name
# Results show entity names like "invoice_inv-002", "non-ar" work fine
# Hyphens in property values don't cause Cypher syntax errors

Verification Steps

1. Run memory extraction: Created memories from conversations with hyphenated terms
2. Check Neo4j database:
docker exec neo4j cypher-shell "MATCH ()-[r]->() RETURN DISTINCT type(r)"
3. Confirm sanitization: Found sub_dash_contractor relationship type